標題: | 半導體雙廠設計方案之評估 Evaluation of Dual-Fab Design Alternatives for Semiconductor Manufacturing |
作者: | 陳曉嚴 Chen, Show-Yen 巫木誠 Wu, Muh-Cherng 工業工程與管理學系 |
關鍵字: | 半導體製造;廠房設計;跨廠生產;群組技術;等候網路模式;Semiconductor manufacturing;Fab design;Cross-fab production;Group technology;Queuing network model |
公開日期: | 2009 |
摘要: | 本研究主要探討在不同規模產出需求下,三種廠房設計方案的表現。三種方案分別為單一大廠房、功能完整雙廠房以及功能分割雙廠房。給定機台種類及數目,三種廠房設計方案主要差異在於廠房內機台設置方式不同。單一大廠房方案將所有相同種類的機台置放一起形成一個工作站,並且將所有工作站設置在單一廠房中。功能完整雙廠房方案將各種類相同的機台分成兩個工作站,兩工作站個別設置在兩個較小的廠房,兩廠皆具備所有種類的工作站,因此定義為功能完整的雙廠房。相反地,功能分割雙廠房則是將相同種類的機台置放一起形成一個工作站,將所有種類的工作站分成兩群,各工作群分別設置到兩小廠,兩廠房內工作站不重複且無法獨立完成產品。本研究將利用等候網路模式來求解廠房設計方案的機台組合,並利用此模式為績效評估工具,評估三種廠房設計方案的表現。在所假設的利潤成本資訊下,實驗結果顯示單一大廠方案適合在小規模產出情境,功能完整雙廠房方案適合在大規模產出情境,而功能分割雙廠房方案則適合中規模產出情境。 This research examines three design alternatives for semiconductor fabs in order to compare their effectiveness under various scales of throughput requirements. The three design alternatives are respectively called single large fab (SL-fab), two functionally-comprehensive fabs (TFC-fab), and two functionally-disjoint fabs (TFD-fab). Given a particular set of machine types and numbers, the three design alternatives use different methods in the decisions of physically locating these machines. In the SL-fab alternative, all machines of each type are placed in a workstation and all workstations are located in a single fab. In the TFC-fab alternative, all machines of each type are grouped into two workstations, each of which as a result is functionally identical and respectively placed in one of two smaller fabs. In contrast, the TFD-fab alternative, all machines of each type are placed in a workstation but all workstations are grouped into two categories, and each category is placed in one of two smaller fabs. Queuing network models are used to design the machine mix for and evaluate the effectiveness of the three design alternatives. Experiment results indicate that the SL-fab alternative is most suitable for small-scale throughput, the TFD-fab alternative is most suitable for mid-scale throughput, while the TFC-fab alternative might be most suitable for very large-scale throughput. Surely, ranges of the three-level scales of throughput depend upon how we make the assumptions about the revenue and cost information. |
URI: | http://140.113.39.130/cdrfb3/record/nctu/#GT079733513 http://hdl.handle.net/11536/45418 |
顯示於類別: | 畢業論文 |